In silico modelling of pharmacological profiles and ADMET properties via advanced machine learning approach
In Silico Research in Biomedicine · 2026 · 2 · 100717
DOI: 10.1016/j.insi.2026.100717 ↗Kurz gefasst
Fachartikel in der Zeitschrift In Silico Research in Biomedicine, 2026. Verfasst von Manju K und 10 weiteren Autor:innen. Die Arbeit nennt 136 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The optimization of pharmacological efficacy alongside favourable Absorption, Distribution, Metabolism, Excretion and Toxicity (ADMET) properties remains a major challenge in contemporary drug discovery, contributing significantly to late-stage clinical attrition. This review provides a comprehensive assessment of recent advancements in in silico methodologies for ADMET prediction and lead optimization, emphasizing the convergence of machine learning (ML), physics-based simulations, and generative artificial intelligence (AI). Modern computational approaches, including graph neural networks (GNNs), multi-task learning (MTL), and transformer-based architectures, have shown capability in capturing complex structure–activity and structure–property similarities, potentially improving prediction for selected endpoints under appropriate validation conditions. The integration of quantum mechanical descriptors derived from Density Functional Theory (DFT) and molecular dynamics (MD) simulations may provide complementary mechanistic information, although gains in predictive reliability require empirical evaluation into molecular reactivity, binding affinity, and conformational behaviour. Emerging strategies such as active learning enable exploration of vast chemical spaces while reducing computational costs and experimental burden. Furthermore, generative frameworks, including Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and diffusion-based models, facilitate the de novo design of compounds with optimized ADMET characteristics. The coupling of AI-driven predictive platforms with automated experimental systems has also enabled self-driving laboratories operating within closed-loop Design–Make–Test–Learn workflows. Collectively, these integrated approaches offer a transformative framework for accelerating drug discovery, improving development efficiency, and enhancing the success rate of therapeutic innovation.
Zitieren
APA
K, M., Baker, S., Kumara, R., S, S., S, A., S, L., K, P. T., Maryam, A., K, R. H., Shayista, H., & S, N. R. (2026). In silico modelling of pharmacological profiles and ADMET properties via advanced machine learning approach. In Silico Research in Biomedicine, 2, 100717. https://doi.org/10.1016/j.insi.2026.100717
Vancouver
K M, Baker S, Kumara R, S S, S A, S L, et al. In silico modelling of pharmacological profiles and ADMET properties via advanced machine learning approach. In Silico Research in Biomedicine. 2026;2:100717. doi:10.1016/j.insi.2026.100717
BibTeX
@article{K2026,
author = {K, Manju and Baker, Syed and Kumara, Ravi and S, Satish and S, Anjani and S, Lakshmi and K, Pavan T and Maryam, Aakifa and K, Ranjini H and Shayista, H. and S, Niranjan Raj},
title = {In silico modelling of pharmacological profiles and ADMET properties via advanced machine learning approach},
journal = {In Silico Research in Biomedicine},
year = {2026},
volume = {2},
pages = {100717},
doi = {10.1016/j.insi.2026.100717}
}
EUVIMED